add support for phrase slop in query language

[saroh]

Breaking changes to the existing query language:

• ~ is now allowed unescaped in the field name

The first commit still allows ~ in words, might be a bit flimsy.
Second one exposes distance for all leaves. Do we need to explicitly fail for leaves that do not handle slop/distance + implement it for those who do ? (Term maybe, I haven't checked)

Not realy sure on the name, slop is hard to understand, distance is maybe a little to vague.

closes #1390

[codecov-commenter]

Codecov Report

Merging #1393 (19f695a) into main (83d0c13) will increase coverage by 0.03%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1393      +/-   ##
==========================================
+ Coverage   94.29%   94.33%   +0.03%     
==========================================
  Files         236      236              
  Lines       43418    43643     +225     
==========================================
+ Hits        40942    41170     +228     
+ Misses       2476     2473       -3     

Impacted Files	Coverage Δ
query-grammar/src/query_grammar.rs	99.67% <100.00%> (+0.01%)	⬆️
query-grammar/src/user_input_ast.rs	97.87% <100.00%> (+0.09%)	⬆️
src/query/phrase_query/phrase_query.rs	91.54% <100.00%> (+0.37%)	⬆️
src/query/query_parser/logical_ast.rs	88.37% <100.00%> (+1.19%)	⬆️
src/query/query_parser/query_parser.rs	94.98% <100.00%> (+0.11%)	⬆️
src/aggregation/agg_result.rs	73.46% <0.00%> (-19.79%)	⬇️
src/schema/facet.rs	89.88% <0.00%> (-0.06%)	⬇️
src/indexer/segment_serializer.rs	98.07% <0.00%> (-0.04%)	⬇️
src/fastfield/mod.rs	94.50% <0.00%> (-0.02%)	⬇️
src/indexer/segment_writer.rs	96.40% <0.00%> (-0.01%)	⬇️
... and 22 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 83d0c13...19f695a. Read the comment docs.

[PSeitz]

For naming I think matching distance is much better than slop

[saroh]

I gave a quick look, I should be able to plug in the fuzzy query for bare words with the ~ operator soon.

[fulmicoton]

@PSeitz @saroh I recommend we stick to slop.

It has the merit of

• being unambiguous with fuzzy matching.
• being in line with Lucene's vocabulary

[saroh]

@fulmicoton @PSeitz so we keep slop in the phrase_query mod but it's ok to use matching distance for the grammar ?

Can we define "it" as being an u8 in the query grammar as it is the type that was chosen for the fuzzy query distance and 255 seems to be good enough of a max distance whatever the Query impl for the time being ?

[fulmicoton]

Right now let's not handle fuzzy query in the grammar, it would break quickwit.
For the naming, I'd stick to slop in the grammar for the moment.

[saroh]

I've got this PR https://github.com/saroh/tantivy/pull/1/files which I can push in here or open later. Adds support for "foo"~1, no changes to the query language.